Surface transfer doping of diamond - Nature

letters to nature Received 5 April; accepted 7 June 2004; doi:10.1038/nature02727. 1. Ko¨hler, J. et al. Magnetic resonance of a single molecular spin. Nature 363, 242–244 (1993). 2. Wrachtrup, J., von Borczyskowski, C., Bernard, J., Orrit, M. & Brown, R. Optical detection of magnetic resonance in a single molecule. Nature 363, 244–245 (1993). 3. Jelezko, F., Gaebel, T., Popa, I., Gruber, A. & Wrachtrup, J. Observation of coherent oscillations in a single electron spin. Phys. Rev. Lett. 92, 076401 (2004). 4. Manassen, Y., Hames, R. J., Demuth, J. E. & Castellano, A. J. Direct observation of the precession of individual paramagnetic spins on oxidized silicon surfaces. Phys. Rev. Lett. 62, 2531–2534 (1989). 5. Durkan, C. & Welland, M. E. Electronic spin detection in molecules using scanning-tunnelingmicroscopy-assisted electron-spin resonance. Appl. Phys. Lett. 80, 458–460 (2002). 6. Tyryshkin, A. M., Lyon, S. A., Astashkin, A. V. & Raitsimring, A. M Electron spin relaxation times of phosphorus donors in silicon. Phys. Rev. B 68, 193207 (2003). 7. Poindexter, E. H. in Semiconductor Interfaces, Microstructure, and Devices (ed. Feng, Z. C.) Ch. 10, 229–256 (Inst. Phys. Publ., Bristol, Philadelphia, 1993). 8. Loss, D. & DiVincenzo, D. P. Quantum computation with quantum dots. Phys. Rev. A. 57, 120–126 (1998). 9. Friesen, M. et al. Practical design and simulation of silicon-based quantum-dot qubits. Phys. Rev. B 67, 121301 (2003). 10. Vrijen, R. et al. Electron-spin-resonance transistors for quantum computing in silicon-germanium heterostructures. Phys Rev. A 62, 012306 (2000). 11. Kane, B. E. A silicon based nuclear spin quantum computer. Nature 393, 133–137 (1998). 12. Kirton, M. J. & Uren, M. J. Noise in solid-state microstructures: a new perspective on individual defects, interface states and low-frequency (1/f) noise. Adv. Phys. 38, 367–468 (1989). 13. Xiao, M., Martin, I. & Jiang, H. W. Probing the spin state of a single electron trap by random telegraph signal. Phys. Rev. Lett. 91, 078301 (2003). 14. Martin, I., Mozyrsky, D. & Jiang, H. W. A scheme for electrical detection of single electron spin resonance. Phys. Rev. Lett. 90, 018301 (2003). 15. Liu, W., Ji, Z., Pfeiffer, L., West, K. W. & Rimberg, A. J. Real-time detection of electron tunnelling in a quantum dot. Nature 423, 422–425 (2003). 16. Kato, Y. et al. Gigahertz electron spin manipulation using voltage-controlled g-tensor modulation. Science 299, 1201–1204 (2003). 17. Yablonovitch, E. et al. Optoelectronic quantum telecommunications based on spins in semiconductors. Proc. IEEE 91, 761–780 (2003). 18. Wallace, W. J. & Silsbee, R. Spin resonance of inversion-layer electrons in silicon. Phys. Rev. B 44, 12964–12968 (1991). 19. Eton, S. S. & Eaton, G. R. Irradiated fused-quartz standard sample for time-domain EPR. J. Magn. Res. Ser. A 102, 354–356 (1993).

Acknowledgements We would like to thank P. M. Lenahan for information regarding the relaxation of spin centres in SiO2, and K. Wang for helping to secure the samples. The work was supported by the Defense Advanced Research Projects Agency and the Defense MicroElectronics Activity. Competing interests statement The authors declare that they have no competing financial interests. Correspondence and requests for materials should be addressed to H.W.J. ([email protected]).

chemical model in which the reduction of hydrated protons in an aqueous surface layer gives rise to a hole accumulation layer6,7. We expect that transfer doping by C60 will open a broad vista of possible semiconductor applications for diamond. The C60 was evaporated onto differently prepared diamond surfaces and onto a silica glass substrate. After each evaporation step the conductance of the respective sample was measured as described below and without breaking the vacuum. The closed symbols in Fig. 1 refer to the conductance of two different homoepitaxial diamond samples with hydrogenated surfaces as a function of C60 coverage in a double-logarithmic plot. The starting point is in both cases the annealed state (see Methods) with a conductance well below 10212 S (leftmost data point). With increasing C60 coverage the conductance rises rapidly by more than six orders of magnitude and saturates at 1026 S for sample D20 and 1025 S for sample D34 at a C60 coverage between 4 and 8 monolayers (ML). Two control experiments (open symbols) were performed with silica glass and with diamond sample D20, this time with an oxidized instead of a hydrogenated surface as the substrate for C60 evaporation. The negative results presented in Fig. 1 confirm that the conductivity necessarily requires the combination of C60 and hydrogenated diamond; the C60 layer alone on an insulating substrate (glass) or in combination with an oxidized diamond surface is not sufficient. We propose that the cause of the observed conductivity is a transfer doping mechanism whereby electrons are transferred from diamond to the C60 layer such that an equal amount of electrons and holes reside on the C60 and diamond side of the interface, respectively. For this to work, the energy level scheme as sketched in Fig. 2a has to be considered. The left-hand part depicts the valence and conduction band edges (VBM and CBM) of diamond, together with the Fermi level E F. On the right-hand side a limited number of occupied and unoccupied molecular orbitals of C60 are schematically sketched with the highest occupied (HOMO) and the lowest unoccupied (LUMO) bracketing the fundamental gap of C60. The common reference level is the vacuum level E VAC. For hydrogenterminated diamond the electron affinity is negative, that is, the CBM lies 1.3 eV above E VAC (refs 9, 10). The electron affinity E VAC 2 E LUMO of C60 is measured to be about 2.7 eV for the C60 molecule11, and this places the Fermi level of C60, which lies between the HOMO and the LUMO, well below the E F of diamond. Hence,

..............................................................

Surface transfer doping of diamond P. Strobel, M. Riedel, J. Ristein & L. Ley Institute for Technical Physics, University of Erlangen, 91054 Erlangen, Germany .............................................................................................................................................................................

The electronic properties of many materials can be controlled by introducing appropriate impurities into the bulk crystal lattice in a process known as doping. In this way, diamond (a well-known insulator) can be transformed into a semiconductor1, and recent progress in thin-film diamond synthesis has sparked interest in the potential applications of semiconducting diamond2,3. However, the high dopant activation energies (in excess of 0.36 eV) and the limitation of donor incorporation to (111) growth facets only have hampered the development of diamond-based devices. Here we report a doping mechanism for diamond, using a method that does not require the introduction of foreign atoms into the diamond lattice. Instead, C60 molecules are evaporated onto the hydrogen-terminated diamond surface, where they induce a subsurface hole accumulation and a significant rise in two-dimensional conductivity. Our observations bear a resemblance to the so-called surface conductivity of diamond4–8 seen when hydrogenated diamond surfaces are exposed to air, and support an electroNATURE | VOL 430 | 22 JULY 2004 | www.nature.com/nature

Figure 1 Conductance of different substrates upon evaporation of C60 in ultrahigh vacuum. Filled circles and squares, hydrogen-terminated diamond, samples D20 and D34, respectively. Open squares, oxygen-terminated diamond; open triangles, silica glass. The leftmost data points correspond to the clean surfaces after mild annealing at about 250 8C. Note the logarithmic scales of the plot. The C60 coverage is given in ML; 1 ML corresponds to an areal density of 1.15 £ 1014 C60 molecules per cm2. The lines through the data points are guides to the eye.

©2004 Nature Publishing Group

439

letters to nature an electron flow from the diamond valence band enters the LUMO of C60 upon contact. The holes left behind set up a space charge region in diamond that results in an upward band bending, as indicated schematically in Fig. 2b. At the same time the Fermi level of the C60 layer moves towards E LUMO and the charge exchange comes to a halt when the two Fermi levels line up. The amount of band bending and the Fermi level shift in C60 are connected through the condition of charge neutrality: the number of holes in diamond equals that of electrons in C60. Combining Poisson’s equation for the space charge density on the diamond side and the Fermi occupation statistics for the diamond and the C60 states, it is a somewhat lengthy but straightforward task to show that the equilibrium charge concentration is entirely determined by the areal density of C60 molecules and the energy separation D ¼ ELUMO;C60 2 VBMdiamond ; which is a fixed quantity independent of any charge exchange and band bending. The carrier density can be estimated from the measured conductivity provided we know the mobility of the respective carriers. For solid C60 the electron mobility at 300 K does not exceed 0.1 cm2 V21 s21 (ref. 12), whereas the mobility of surface-near holes in diamond lies typically between 50 and 100 cm2 V21 s21 (ref. 13). Hence the conductivity must be ascribed to the holes in diamond. Adopting an average value of m < 70 cm2 V21 s21 for the hole mobility we translate the twodimensional conductivity j A ¼ emn (where e is the elementary charge) into the two-dimensional hole density n (right-hand scale in Fig. 3). For sample D34 we estimate a maximum sheet carrier concentration of 1.5 £ 1012 cm22 for a C60 coverage of 7 ML. The simulation based on Poisson’s equation and Fermi statistics yields a corresponding value of 0.24 eV for D. With this value of D the concentration of transferred electrons and the ensuing conductivity have been calculated as a function of C60 coverage. The result, compared in Fig. 3 with the measurements for sample D34, is far off the mark. With decreasing coverage, the actual conductance drops precipitously by six orders of magnitude between about 2 and 0.1 ML of C60, whereas the calculated value decreases by a mere factor of ten over the same range. The straightforward solution, namely to attribute the experimental drop to a corresponding drop in mobility, is unreasonable. If anything, mobilities tend to increase with decreasing carrier density. This leaves an increase in D as the only cause for the discrepancy in the framework of the model. In Fig. 4 we have calculated values for D as a function of C60 coverage such that they reproduce the conductivity data of sample D34. A

very similar relationship, albeit with fewer points, holds for D20. In essence, Fig. 4 says that it costs 0.24 eV to transfer an electron from the diamond valence band maximum into the LUMO of a C60 layer several ML thick, whereas the same transfer costs 1.55 eV if it takes place into an isolated C60 molecule. Phrased differently, the electron affinity increases by 1.31 eV when going from a single C60 molecule to a C60 solid. The electron affinity is, strictly speaking, the difference of two total energies, namely that of the system in its N-electron ground state E(N) and that of the N þ 1 electron system also in its ground state, E(N þ 1). The difference E(N) 2 E(N þ 1) is a true manyelectron quantity that depends on the size of the system. For the sake of argument it is convenient to split the many-electron expression into a difference of one electron energies (that is, E VAC 2 E LUMO) and a many-body correction that is usually termed the effective correlation energy U. This energy involves the Coulomb interaction of the additional electron brought into the lowest accessible electronic state of the system with all the other electrons already present. As such, it depends on, and is reduced by, the relaxation of the electrons already present, and this relaxation in turn depends strongly on the polarizability of the system. Consequently, the effective correlation energy is significantly lower for an electron brought to the conduction band minimum of C60 in the solid phase than into the LUMO of an isolated C60 molecule. For such a molecule, theory gives values between 2.7 and 3.1 eV for the correlation energy, whereas an experimental estimate yields a slightly larger value of 3.3 eV (ref. 14 and references therein). The same quantity for the solid state was found between 0.8 and 1.3 eV theoretically and between 1.4 and 1.6 eV experimentally14. The difference dU < 1.8 ^ 0.2 eV in effective correlation energy is in reasonable agreement with the change of 1.31 eV in D derived above. In fact, electrons and holes in the transfer doping process have only to be separated across the width of the diamond space charge layer plus the C60 layer thickness. This leaves a finite—though hard to specify—Coulomb energy by which the ‘band offset’ energy D should be smaller than the difference between the diamond ionization potential and the C60 electron affinity. The remaining Coulomb energy is expected to decrease with increasing C60 layer thickness, and the decrease in D should indeed be smaller than dU by this difference.

e–

Figure 2 Schematic representation of the surface transfer doping of intrinsic diamond by C60. a, Level scheme before electron transfer from the diamond side with the higher-lying E F to the C60 side with the lower-lying E F. b, Band bending in chemical equilibrium. An equal amount of holes in diamond and electrons in the lowest (normally, that is for neutral C60 at zero temperature) unoccupied molecular orbital (LUMO) of the C60 layer with a concomitant electrostatic potential on the diamond side is established. 440

Figure 3 Comparison between experiment and simulation of the two-dimensional conductivity of hydrogen-terminated diamond as a function of C60 coverage. The simulation is based on a quantitative evaluation of the surface transfer doping model illustrated in Fig. 2. A ‘band offset’ of D ¼ 0.24 eV, independent of C60 coverage, has been assumed as the only free parameter of the simulation. The sheet conductivity (lefthand ordinate) and the areal hole density (right-hand ordinate) are proportional to each other with the hole mobility of 70 cm2 V21 s21 and the elementary charge as proportionality constants.


NATURE | VOL 430 | 22 JULY 2004 | www.nature.com/nature

letters to nature The fact that no conductivity is observed for an oxidized diamond surface stresses the importance of the low ionization energy of diamond in conjunction with the rather high electron affinity of C60. Removal of hydrogen raises the ionization potential by 1.6 eV and oxidation by more than 3 eV (ref. 10), with the result that transfer doping is energetically not possible and conductivity no longer observed. As a final control experiment we used hydrogenated but nitrogen-doped synthetic type Ib diamond as substrate and observed no conductivity (j A # 1013 S per square) upon evaporation of C60. The electron affinity as a pure surface property does not depend on substrate doping. Thus the same transfer of electrons to C60 as for intrinsic diamond is expected. However, nitrogen forms a deep donor in diamond and all holes that are created as a result of transfer doping recombine with electrons from these compensating defects. The charge exchange results in a positive space charge (and a corresponding upward band bending) in the diamond; this time, however, the positive space charge is localized as ionized Nþ defects and not present as mobile holes. On the C60 side we expect no difference compared to the experiment on intrinsic diamond; in particular, an electron transfer into the LUMO is also expected to take place as well. The low conductance of the sample confirms the negligible mobility of those electrons and is an indirect proof that the assignment of the observed conductivity to holes in diamond is correct. The results presented above bear a very close resemblance to the so-called surface conductivity of diamond. Discussions on the origin of this type of surface conductivity, which is observed when hydrogenated diamond surfaces are exposed to air, have generated controversy. The suggestions range from hydrogeninduced acceptor states4, eventually in combination with compensating surface defects5, to an electrochemical model in which the reduction of hydrated protons in an aqueous surface layer gives rise to the hole accumulation layer6,7. The electrochemical model met with a certain amount of scepticism because a charge transfer from diamond to an adsorbate across the interface was thought unrealistic. In view of the results presented here it can no longer be doubted that the electrochemical model for the air induced surface conductivity is basically correct, because it relies on a mechanism similar to the one described here. From a technical point of view the new transfer doping by C60 is likely to replace the air-induced surface conductivity for most of the devices that have been realized on the basis of the surface conductivity of diamond8,15. The ease of fabrication, higher reliability and reproducibility of the C60 doping are assets that will ultimately improve the device characteristics and could bring new concepts to fruition. A

Methods Sample preparation and characterization For the experiments the following types of diamond samples have been used. The main work was performed on undoped, homoepitaxial diamond layers with (100) orientation (samples D20 and D34) that were grown by microwave-assisted vapour phase epitaxy to a thickness of about 0.5 mm on nitrogen-doped (N concentration 1019 cm23), nonconducting, synthetic high-pressure, high-temperature diamond substrates. A natural, boron-doped and therefore conducting, IIa diamond was used for photoelectron spectroscopy of the C60 layer, and finally a nitrogen-doped substrate (without an intrinsic epilayer on top) was used for reference measurements. All samples, except those for which the converse is explicitly stated, were plasma-hydrogenated to saturation in a separate preparation chamber before the experiments. The experiments were performed in interconnected ultrahigh vacuum chambers with a pressure below 1029 mbar. Owing to their contact with air, the as-prepared samples exhibited the well-known atmosphere-induced surface conductivity, which had to be removed in a first preparation step. To this end, the diamond samples were initially annealed in UHV at temperatures between 200 and 300 8C. These temperatures are known to remove all residual surface conductivity while leaving the hydrogen termination intact16. This was confirmed in situ by total photoelectron yield spectra which exhibit the signature characteristic for hydrogen-terminated surfaces with negative electron affinity17. The C60 was evaporated onto the diamond surfaces and onto the silica glass reference substrate at a rate of typically 0.1 ML min21. The nature of the evaporated layer as C60 was confirmed on the basis of its valence band spectrum, the characteristic electron energy loss spectrum, and the shift of 1.3 eV of its C1 s line relative to the C1 s binding energy of diamond—all measured by photoelectron spectroscopy18. Photoelectron spectroscopy was also used to determine the C60 layer thickness from the attenuation of the diamond signal.

Conductivity measurements After each evaporation step of C60, two gold-coated, spring-loaded tips of r ¼ 0.05 mm radius at a centre distance of D ¼ 2 mm were pressed onto the sample surface; the conductance of the different specimens was then determined in situ by measuring the I 2 V characteristics between 210 and þ10 V. The ohmic behaviour was verified for each measurement and the conductance derived by a linear regression. The conductance is (aside from a geometric factor g of the order of unity) equal to the two-dimensional surface conductivity j A. For the tip arrangement used here, g ¼ lnðD=r 2 1Þ=p ¼ 1:17; and in view of the large variation of conductivities observed in our experiments, conductance and two-dimensional conductivity could be used as synonyms throughout this paper. Received 8 April; accepted 8 June 2004; doi:10.1038/nature02751. 1. Nebel, C. E. & Ristein, J. (eds) Thin-Film Diamond I Monogr. Ser. Semiconductors and Semimetals Vol. 76, 145–259 (Elsevier, New York, 2003). 2. Nebel, C. E. & Ristein, J. (eds) Thin-Film Diamond II Monogr. Ser. Semiconductors and Semimetals Vol. 77, 97–338 (Elsevier, New York, 2004). 3. Isberg, J. et al. High carrier mobility on single-crystal plasma deposited diamond. Science 297, 1670–1672 (2002). 4. Denisenko, A. et al. Hypothesis on the conductivity mechanism in hydrogen terminated diamond. Diamond Relat. Mater. 9, 1138–1142 (2000). 5. Takeuchi, D., Yamanaka, S. & Okushi, H. Schottky junction properties of the high conductivity layer of diamond. Diamond Relat. Mater. 11, 355–358 (2002). 6. Ri, S. G. et al. Formation mechanism of p-type surface conductive layer on deposited diamond films. Jpn. J. Appl. Phys. 34, 5550–5555 (1995). 7. Maier, F., Riedel, M., Mantel, B., Ristein, J. & Ley, L. The origin of surface conductivity in diamond. Phys. Rev. Lett. 85, 3472–3475 (2000). 8. Kawarada, H. Hydrogen terminated diamond surfaces and interfaces. Surf. Sci. Rep. 26, 205–259 (1996). 9. Cui, J. B., Ristein, J. & Ley, L. The electron affinity of the bare and hydrogen covered single crystal diamond (111) surface. Phys. Rev. Lett. 81, 429–432 (1998). 10. Maier, F., Ristein, J. & Ley, L. Electron affinity of plasma-hydrogenated and chemically oxidized diamond (100) surfaces. Phys. Rev. B 64, 165411 (2001). 11. Yang, S. H., Petiette, C. L., Conceicao, J., Cheshnovsky, O. & Smalley, R. F. UPS of buckminsterfullerene and other large clusters of carbon. Chem. Phys. Lett. 139, 233–238 (1987). 12. Haddon, R. C. et al. C60 thin film transistors. Appl. Phys. Lett. 67, 121–123 (1995). 13. Hayashi, K. et al. Investigation of the effect of hydrogen on electrical and optical properties in chemical vapor deposited homoepitaxial diamond films. J. Appl. Phys. 81, 744–753 (1997). 14. Gunnarson, O. Superconductivity in fullerides. Rev. Mod. Phys. 69, 575–606 (1997). 15. Gluche, P., Aleksov, A., Vescan, A., Ebert, W. & Kohn, E. Diamond surface-channel FETstructure with 200 V breakdown voltage. IEEE Electr. Dev. Lett. 18, 547549 (1997). 16. Riedel, M., Ristein, J. & Ley, L. Recovery of surface conductivity of H-terminated diamond after thermal annealing in vacuum. Phys. Rev. B 69, 125338 (2004). 17. Bandis, C. & Pate, B. B. Photoelectric emission from negative-electron-affinity diamond (111) surfaces: Exciton breakup versus conduction-band emission. Phys. Rev. B 52, 12056–12071 (1995). 18. Golden, M. S. et al. The electronic structure of fullerenes and fullerene compounds from high energy spectroscopy. J. Phys. Condens. Matter 7, 8219–8247 (1995).

Acknowledgements We acknowledge A. Hirsch for the gift of C60 and fruitful discussion. Partial financial support by the Deutsche Forschungsgemeinschaft is also gratefully acknowledged.

Figure 4 Reduction of the ‘band offset’ D (see Fig. 2) between the diamond valence band maximum and the C60 LUMO with increasing C60 coverage. This reduction reflects the increase of the electron affinity of C60 when going from isolated molecules to the molecular solid. NATURE | VOL 430 | 22 JULY 2004 | www.nature.com/nature

Competing interests statement The authors declare that they have no competing financial interests. Correspondence and requests for materials should be addressed to J.R. ([email protected]).


441